Inspiration
We love AI, but it isn't very accessible, and we wanted to change that.
What it does
You download the AI model to your computer and it performs the calculations with that downloaded model only on your computer, with complete privacy and data safety!
How we built it
We found some this project called llama.c, based on llama.cpp code, that was an 700 liner pure c file (no dependencies) that led to a TypeScript and Javascript implementation, which we removed Node.js dependencies for to make it run on web. Once done, we remade the website GUI.
Challenges we ran into
The larger models use a specific architecture (.gguf and .ggmc) that would take days to write up javascript code for from scratch. We never got it to work and limited the scope to smaller models.
Accomplishments that we're proud of
We created a running website with an intuitive AI chatbot.
What we learned
We learned a lot! So much AI is out there!!!
What's next for webgpt
Supporting larger llms, though maybe throuch llama.c and wasm for low-level speed ups, for supporting more models.
Log in or sign up for Devpost to join the conversation.